A New Approach to Identify Reactive and Neoplastic Neoformation Bone Based on CT Radiomics: A Retrospective Multicenter Study

Background: To identify neoplastic and reactive neoformation bone by quantitative CT radiomics analysis, and to further validate the performance of the model by the diagnosis of osteosarcoma (OS) and chondrosarcoma (CS). Methods: A total of 478 bone tumor patients were retrospectively enrolled from 5 hospitals. Crucial radiomics features were selected and employed to construct radiomics signature by Least absolute shrinkage and selection operator (LASSO) regression. Then radiomics signature and clinical characteristics were combined to construct the model identifying neoplastic and reactive neoformation bone (ModelRN) in the training set (n = 300), and then validated in the validation set (n = 178). Furthermore, a new model (ModelOC) incorporating the clinical characteristics and the comprehensive scores produced by ModelRN was constructed to differentiate OS and CS (training set: n = 103; validation set: n = 41). The performance of the model was evaluated with respect to discrimination, calibration, and clinical usefulness.   Findings: ModelRN incorporating radiomics signature, tumor position and course of disease achieved promising results with high identification accuracy (AUC: 0.9319 in the training set and 0.9350 in the validation set). ModelOC was built incorporating alkaline phosphatase with the comprehensive score produced by ModelRN. ModelOC also showed satisfying performance in the training (AUC: 0.8752) and validation (AUC: 0.9032) sets. Decision curve analysis showed the clinical usefulness of the model. Interpretation: This study suggested that CT-based radiomics can help radiologists to identify neoplastic and reactive neoformation bone, thereby more effectively distinguishing osteosarcoma from chondrosarcoma than current clinical images diagnosis. Funding Statement: This paper is supported by the National Natural Science Foundation of China (Grant Nos. 81871510, 81871511), Ministry of Science and Technology of China under Grant (No.2017YFA0205200, 2017YFA0700401, 2016YFA0100902, 2016YFC0103803, 2016YFA0201401, 2016YFC0103702). Declaration of Interests: The authors declare no potential conflicts of interest. Ethics Approval Statement: Ethical approval was obtained from the institutional review board (IRB) of the Third Affiliated Hospital of Southern Medical University, and the need for informed consent from the patients was waived.